Server consolidation with migration control for virtualized data centers

Abstract

Virtualization has become a key technology for simplifying service management and reducing energy costs in data centers. One of the challenges faced by data centers is to decide when, how, and which virtual machines (VMs) have to be consolidated into a single physical server. Server consolidation involves VM migration, which has a direct impact on service response time. Most of the existing solutions for server consolidation rely on eager migrations, which try to minimize the number of physical servers running VMs. These solutions generate unnecessary migrations due to unpredictable workloads that require VM resizing. This paper proposes an LP formulation and heuristics to control VM migration, which prioritize virtual machines with steady capacity. We performed experiments using TU-Berlin and Google data center workloads to compare our migration control strategy against existing eager-migration-based solutions. We observed that avoiding migration of VMs with steady capacity reduces the number of migrations with minimal penalty in the number of physical servers.

Introduction

Virtualization has become a fundamental asset in several areas of Computer Science. Despite being an old concept (initially used by IBM 370 mainframes [1]), virtualization assists current organizations in dealing with problems such as unpredicted demand of computing resources, high management and energy costs, and security. As a consequence, there has been an increasing number of commercial and open-source software products (e.g. [2], [3], [4], [5]) and investments of major hardware companies to provide better support for virtualization in their products (e.g. [6], [7]).

One of the benefits of virtualization is the possibility of gathering several virtual machines (VMs) into a single physical server. This process, known as server consolidation, is used by data centers to increase resource utilization and reduce electric power consumption costs. Server consolidation is particularly important when user workloads are unpredictable and need to be revisited periodically. Whenever a user demand changes, VMs can be resized and migrated to other physical servers if necessary.

Our research hypothesis is that a more conservative approach can be used to provide steadier performance guarantees. Our proposal extends current server consolidation solutions by including constraints to define that virtual machines with steady usage are not migrated and virtual machines with variable capacity can be migrated to reduce the number of required physical servers. We call this new approach as dynamic consolidation with migration control. Thus, the contributions of this paper are (i) an LP formulation and heuristics to control VM migration, which prioritize virtual machines with steady capacity and (ii) an extensive evaluation based on TU-Berlin and Google data center workloads that compares our approach with existing ones and corroborates our research hypothesis.

The obtained results are encouraging, since they show that a more conservative migration approach can reduce the number of migrations with minimal penalty in the number of physical servers compared to eager-migration-based solutions [8], [9], [10], [11], [12], [13], [14], [15]. This research thus has a direct impact on service response times and cost saving in data centers.